Parameter Constrained Neural Network

Model overview

Parameter constrained neural network classifier using PyTorch with configurable hidden layers to predict taxonomy IDs from product tags. Each taxonomy path is encoded as a p-adic integer and the model minimizes p-adic distance.

5 CV folds

11.68% Mean accuracy

0.0854 Mean F1

0.55407558 Mean p-adic loss

Total train samples: 54,236
Total test samples: 13,559

Key hyperparameters

Prime base(s): 79
Hidden layer sizes: 27
Max tags used: 32

Cross-validation results

Fold	Accuracy	F1	P-adic loss (mean)	Details
0	12.89%	0.0960	0.547302	View details →
1	11.46%	0.0843	0.569259	View details →
2	11.55%	0.0810	0.554561	View details →
3	11.05%	0.0784	0.546161	View details →
4	11.46%	0.0872	0.553095	View details →